Skip to main content
Every image you upload to Studio is a piece of evidence that Gemini uses to understand your product. The front and back product images are the source of truth for all five poses — they determine silhouette, color, fabric, construction details, and everything that must appear correctly in the final catalog images. Optional references add depth, texture, and creative direction without overriding the product truth.

​
File requirements

​
Standard garment references (non-saree)

Front product image

Required. The primary source of truth for silhouette, color, print, neckline, sleeves, and all front construction details. Every pose uses this image to anchor the garment’s appearance.

Back product image

Required. Used to generate Pose 3 (Full Back View) accurately. Gemini uses it to document the real back neckline, closure, and construction — details that would otherwise be guessed.

Fabric / pattern detail

Optional. A close-up of the fabric weave, print motif, embroidery, or surface texture. Gemini uses this specifically for Pose 5 (Zoomed-In Face & Product Highlight) to capture fine surface detail.

Bottom wear / farshi

Optional. Trousers, skirt, or farshi — captures cut, volume, and print. Locking bottom wear ensures the full outfit is generated consistently across all poses.

Mannequin / flat-lay shot

Optional. A dress form or flat-lay that shows the worn shape and drape. Helps Gemini understand how the garment falls when worn rather than relying solely on flat product imagery.

Additional product photo

Optional. Any extra angle or detail — side seams, labels, trims — that the front and back shots don’t capture.
Style references guide scene and lighting only. Gemini will never use them as product truth — the front and back product images always take priority. Upload style references to suggest a mood, backdrop color, or editorial feel, not to change the garment itself.

​
Saree references

Sarees require a more comprehensive evidence set because Gemini must reconstruct the body fabric, pallu, borders, blouse, and drape physics separately before building the pose plan.
If Gemini detects a saree from a generic category setting, it automatically promotes your front and back references to the front drape and back drape roles and asks you to confirm or upload the missing saree-specific evidence before generation is unlocked.

​
Mapping legacy evidence to saree roles

If you uploaded a fabric close-up or additional product photo before switching to the saree category, Studio shows a Map available product evidence panel. Use the Use as pallu or Use as body detail buttons to explicitly reclassify one image to the correct region. The image moves to the chosen role — it is not silently used for both.
A pallu image must show the pallu fully opened out, not just a fabric close-up. Only reclassify an image to the Pallu spread role when it genuinely shows the pallu artwork and edges.

​
Style references

Upload up to three style reference images to guide scene, mood, and lighting direction. These are sent to Gemini as creative context, not as product evidence.
  • Maximum three style references per shoot
  • Each style reference follows the same file requirements: PNG, JPEG, or WebP, maximum 20 MB
  • Style references affect background, lighting, composition, and editorial feel only
  • They never override the garment’s color, silhouette, fabric, or any detail derived from the front and back product images

​
Model face lock

Upload a single clear face photo to lock the model’s identity across all five poses. When a model face reference is present, GPT Image 2 reproduces that face as closely as photographically possible for every frame instead of generating its own consistent model. Leave this field empty to let the shoot design one consistent model automatically.

​
Replacing a reference

To swap any reference image, hover over its preview card and click Replace, then select a new file. You can also drag and drop a new file directly onto the card. Replacing any reference — including style references and the model face lock — immediately marks the current analysis as stale. Generation is locked until Gemini re-analyzes with the updated references. Re-analysis starts automatically a moment after you finish uploading.
Generation cannot start without both the front and back product images for standard garments, or without the four required saree references (front drape, back drape, pallu spread, and body detail) for sarees. The Generate button remains disabled until all required references are uploaded and the analysis is current.

​
30-day analysis cache

If you upload images that are byte-for-byte identical to a previous session — same files, same category, same model identity, same background style, same product details — Studio reuses the existing analysis result instead of making a new Gemini call. Any change to any reference or setting bypasses the cache and triggers a fresh analysis.